ASCEND: A randomized controlled trial of titration strategies for vagus nerve stimulation in drug-resistant epilepsy

Stimulation


a b s t r a c t
Vagus Nerve Stimulation (VNS) therapy is widely understood to provide clinically meaningful improvements in seizure control to patients with drug-resistant epilepsy, and has been a staple in the clinical armamentaria available to epileptologists for over 25 years. Despite the long history of evidence-based reviews by neurology professional societies, there is still evidence of a practice gap in VNS titration and dosing that aims to maximize clinical benefit. Recent retrospective analyses have strongly argued for a more consistent application of a population-wide target dose of VNS, and further argued the importance of quickly achieving this target dose to hasten the onset of clinical benefits; however, these analyses failed to provide evidence for practical implementation. Herein, we describe a randomized controlled trial assessing the impact of titrating VNS according to three different protocols to achieve the target dose of 1.5 mA at 500lsec, for a 20-Hz signal frequency. The study was registered as NCT02385526 on March 11, 2015. Sixty-two patients were randomized into treatment groups that followed different titration protocols. One protocol (Group A) was designed to align with currently accepted professional guidance for VNS titration and the manufacturer's labeling for VNS in epilepsy (Heck et al., 2002), while the other two protocols were derived from VNS applications in other therapeutic areas. Group A participants were most likely to achieve the target dose parameters in 12 weeks or less (81.8%), with a median time-until-achievement of the target dose of 8.1 weeks, while less than 60% of patients in other groups were able to achieve the same endpoint. Participants in all groups experienced low levels of transient tolerability concerns and adverse events, suggesting titration to the target dose in 12 weeks or less following the Group A protocol is generally acceptable to most patients. These findings indicate that patients receiving VNS for epilepsy can achieve the manufacturer-recommended dose range in 12 weeks or less. A wider implementation of the approach will likely improve the clinical impact of VNS on seizure control and prevent undertreatment.

Introduction
Almost 30 years ago (1994), VNS therapy ''launched the modern era of neurostimulation" [2], and subsequently grew from ''an interesting and novel therapy for intractable epilepsy" [3] to the most widely used form of neurostimulation for treating drugresistant epilepsy (DRE). Despite being deemed ''effective and safe, based on a preponderance of Class I evidence" nearly 25 years ago [4], and being the only neurostimulation modality that was reviewed twice (1997 and 1999) by the AAN Therapeutics and Technology Assessment Subcommittee [3,4] and once by the Guideline Development Subcommittee of the American Academy of Neurology [5], we still do not have the specific guidelines for its optimal clinical implementation, leading to possible underperformance against its potential and underutilization.
One of the first systematic attempts to provide some scientific basis and recommendations for use of the VNS settings was undertaken by Heck et al. [1]. This review gleaned several cardinal points that stood the test of time: duty cycles (DC) of 30 30 Hz, at output currents (OC) from 0.25 to 3.5 mA, are safe and effective; efficacy of VNS increases over the first year; non-responders to initial settings may respond to incremental increases in OC and DC; DC 50% seem safe and effective; side effects (hoarseness, cough, and throat discomfort) are primarily a function of OC and pulse width (PW) and usually respond to their adjustments. The 2013 AAN guideline [5] was focused on the evidence for patient selection, expected seizure and non-seizure outcomes, and dosing or titration. A determination was that ''optimal VNS settings are still unknown, and the evidence is insufficient to support a recommendation for the use of standard stimulation vs rapid stimulation to reduce seizure occurrence (Level U)". In a clinical contextualization of their determination, the authors caution about the importance of carefully considering the impact of rapid cycling on a battery life [5].
The most comprehensive attempts to identify optimal VNS parameters were undertaken very recently. Fahoum et al (2022) aspired to identify a target dose for VNS therapy for DRE [6]. The authors were able to infer a population level target OC (1.61 mA) and DC (17.1%) associated with the highest likelihood of favorable response based on a statistical analysis of 1178 pooled subjects. The closest available settings are OC of 1.625 mA and DC of 16% (30 s ON/3m OFF). Since the analysis revealed that the patients treated longer with a VNS were more likely to respond to the treatment, independently from dosing effects, an assumption was that patients who have been continually underdosed may still benefit from achieving the target dose. Thus, VNS outcomes may potentially be improved with a proper implementation of evidencebased dosing and titration guidelines -that remain poorly defined.
Tzadok et al (2022) retrospectively analyzed the aggregated data from 12 clinical VNS studies sponsored by LivaNova from 1990 to 2017 (N = 1178; the same cohort as in Fahoum et al, 2022) and inferred that the fast titrated patients (<3 M) attain clinical benefits faster -irrespective of their age [7]. However, two significant concerns ensued: Only about 1 in 10 studied patients were afforded this opportunity resulting from routine clinical practice, and a rapid titration protocol which would improve its tolerability profile (and likely increase a comfort of practicing physicians) remains unknown. Curiously, rapid titration was even less prevalent in children (8% vs. 12% of adults) who are more tolerant to the VNS treatment. In fact, about a half of patients exposed to a slower titration of VNS required over 12 months to attain a target dose [7].
It seems that the last two retrospective analyses have identified a promising destination (a population level target OC and DC [6]), and implied that arriving at it sooner appears clinically beneficial (fast titrated patients attain clinical benefits faster [7]), yet a precise journey (a specific VNS titration protocol) remained elusive. While it was designed and executed before these pivotal manuscripts came to exist, the E-40 (''ASCEND") study was designed to fill this gap in practical understanding of the pace of optimal VNS therapy implementation.

Methods and analysis
The E-40 ''ASCEND" study was designed to determine whether alternate VNS therapy parameters programmed during the titration period can safely accelerate adaptation of VNS tolerance, and thereby shorten the time required to achieve a target dose within a specified time frame. The study was conducted between March 2015 and October 2016 (results posted to clinicaltrials.gov in July 2018). This post-market double-blind randomized trial studied VNS titration in a population of subjects 12 years old with refractory focal-onset seizures treated with VNS therapy. Participants were followed up for 12 weeks after implantation to determine the likelihood that they would reach the labeling-defined target VNS settings after following one of three different titration protocols. Safety and clinical outcomes were collected over the course of the study.
The ASCEND study was conducted in compliance with ISO 14155, FDA 21 CFR Parts 50, 54, 56, and 812, the Declaration of Helsinki, and ICH E6 Good Clinical Practice (GCP). The clinical investigation plan was submitted and approved by the Western Institutional Review Board. The study is registered as NCT02385526.

Population
ASCEND aimed to enroll 60 participants indicated to be treated with VNS therapy. The decision to treat with VNS must have been made independently of and prior to participation in the study. Patients of age 12 or older were eligible for enrollment, and the patient or their caregiver must have been willing and able to give accurate side-effect reports, global impressions data, and complete study questionnaires and surveys with minimal assistance throughout the study. Patients were not eligible for enrollment if they used or expected to use short-wave, microwave, or therapeutic ultrasound diathermy, or expected to undergo an MRI with a body coil for transmission of RF during the clinical study. Patients who had progressive neurological conditions were also excluded.
Sixty-seven (67) patients were screened for the ASCEND trial, with 62 patients eventually being enrolled and randomized into the intent-to-treat population (Fig. 1). These participants were randomized into 3 groups, A, B, and C, that defined what titration strategy each participant received (Table 1). Fifty-seven (57) patients completed all study visits per protocol.

Primary outcome measures
The data collected in this study were not used for specific seizure-related efficacy determinations. The primary efficacy outcome measure, achiever rate, was defined as the proportion of patients in each group who reached the target parameters during any of the post-baseline visits. Safety was summarized based on incidence of adverse events (AEs), serious adverse events (SAEs), and unanticipated adverse device effects (UADEs). AEs that occurred during the study were coded and tabulated utilizing Med-DRA version 17.1. All adverse events were followed by the site investigator until resolution, and the study sponsor was notified within 24 hours of detection of any SAE or UADE. Site investigators were also responsible for reporting these events to their respective IRB in accordance with the institution's regulations.

Secondary outcome measures
Secondary objectives of the ASCEND study included: Evaluation of the time required in each VNS therapy titration group to achieve 1.5 mA, 20 Hz, 500 ls, and DC comprised of 30 s ON / 5 min OFF. Evaluation of differences in tolerability of VNS therapy in the three titration groups Evaluation and comparison of AEs that may occur in each of the titration groups Evaluation of the VNS therapy effectiveness for non-seizure outcomes between the three VNS therapy titration groups using the following scales: CGI-I, PGI [8], and QOLIE-31/QOLIE-48-AD [9][10][11].
While collection of the primary outcome measures of achiever rate and adverse events were collected at all study visits, collection of the CGI-I, PGI, and QOLIE scales were collected only at baseline and 12 weeks after implantation. Patient satisfaction surveys were also collected as exploratory outcome measures.

Group definition and VNS titration
Patients were randomized to one of three groups by a thirdparty statistical program using a predefined randomization schedule. Participants in each group were titrated according to a specific protocol, but all participants were required to achieve the following settings to be considered an achiever of the target dose: 1.5 mA, 20 Hz, 500 ls, and DC comprised of 30 s ON / 5 min OFF (10%). The titration schedule for each group is described briefly in Table 1.
Participants in Group A were set to parameters identical to the final target settings with exception of OC, which started at 0.25 mA. At each study visit, participants had their OC increased to the maximum tolerable OC at that visit. Participants in Groups B and C followed a similar titration schedule; however, they started at a lower signal frequency (SF) of 5 Hz and lower PW of 130 ls, and in Group C patients also had a variant DC of 14 seconds ON and 1.1 minutes OFF (21%). In these groups, patients were required to achieve an OC of 2.5 mA before titrating these other VNS parameters to achieve the target settings for SF, PW, and DC. A higher interim target OC of 2.5 mA was intended to account for the reduced pulse width (from 500lsec to 130lsec).
While the Group A titration algorithm was largely consistent with common VNS therapy titration protocols in epilepsy [1], titration in Groups B and C were variants under investigation. The rationale behind the starting settings in these groups hinged on anecdotal reports that lower SFs and PW are more tolerable for patients. The initial DC setting for Group C was consistent with investigational VNS settings for heart failure and was supportive of rapid titration in that patient population [12].

Statistical analysis
The sample size estimate for this study was not based on formal statistical power calculations. This was an exploratory study to gather general information on standard and alternate titration methods. From previous clinical experiences, a sample of 60 participants (20 per group) was considered sufficient to provide precision of estimates of the assessments, even if comparisons between groups were not sufficiently powered. The study defined 3 populations: The intention-to-treat (ITT) population, the safety population, and the completer population. Most study outcomes were assessed on the ITT population (n = 62), which were participants enrolled, were randomized to a treatment group, and attended at least one successful adjustment of the VNS therapy system parameters. The safety population included all participants implanted with a VNS device (n = 63). The completer population completed every study activity per protocol and is not examined herein.
The primary outcome of achiever rate (AR) was defined as the proportion of ITT participants in each VNS therapy titration group that reached the target parameters during any of the post-baseline visits. A ''non-achiever" was any patient where the target parameters were NOT reached at any post-baseline visit assessment. Achiever rates and associated 95% CI at weeks 2, 4, 6, 8, 10, and 12 were calculated and analyzed for groupwise trends. Cumulative achiever rates (CAR), defined as the proportion of achievers up to a particular visit, and their associated 95% CIs were also calculated at week 2, 4, 6, 8, 10 and 12.  Table 1 Titration parameters for the ASCEND study. Each participant was randomized to one of three arms, with each arm receiving a different protocol for titration. Group A participants were initially programmed to settings consistent with the target settings, with the exception of output current (which was increased steadily over time). Groups B and C were programmed to alternate starting settings hypothesized to improve tolerability during the titration process, namely a lower signal frequency and pulse width. In Group C, participants also experience more frequent repetitions of VNS due to short ON and OFF times. In Groups B and C, participants were required to achieve an interim output current of 2.5 mA to account for the reduced total charge associated with the shorter pulse width (130 ls). Secondary outcome measures were analyzed within the ITT population. Statistical comparisons were conducted to test the differences between treatment Group A and Group B, and Group A and Group C, but these tests were considered exploratory as the study was not sufficiently powered for these tests prospectively.
Time-until-achievement (TUA) is based on achieving target parameter settings (1.5 mA, 20 Hz, 500 lsec, and DC comprised of 30 sec-ON: 5 min-OFF). The TUA was computed as the date a participant achieved the target parameter settings and sent home the first time minus the Implant date (in weeks). Kaplan-Meier analysis was used to compare the response experiences of patients in the three treatment arms. Patients who did not achieve the target settings by the end of the study were censored; on the last visit date the achievement assessment was recorded. Patients who discontinued early were censored; on the last visit date a programming event was assessed. Kaplan-Meier survival curves and median estimates and the associated 95% CIs showing time-tofirst observed achievement of the target settings are summarized by treatment arm. In addition, estimated population ARs and their associated 95% CIs at weeks 2, 4, 6, 8, 10, and 12 are estimated based on the Kaplan-Meier model for each treatment arm. The Quality of Life in Epilepsy Inventory (QOLIE-31) [9] consists of 31 questions divided into seven multi-item subscales: emotional well-being, social functioning, energy/fatigue, cognitive functioning, seizure worry, medication effects, and overall quality of life. A QOLIE-31 ''overall score" is obtained using a weighted average of the multi-arm scale scores and was computed at baseline and week 12 for each patient. Descriptive statistics of the overall score (mean, SD, median, Q1, Q3, min, max) were calculated for each treatment group at baseline and at week 12. An ANCOVA was performed to analyze change from baseline for Group A vs. Group B and Group A vs. Group C with the baseline score as a dependent variable in the model. The least square means of the differences of Group B vs. Group A and Group C vs. Group A was presented along with associated 95% CI by visit. Change from baseline for the ''overall score" for Group A vs. Group B and Group A vs. Group C were compared by means of least square means and 95% CIs. The QOLIE-AD 48 scale assesses Quality of Life in Epilepsy for patients between 12 years and 17 years. Only one patient completed this scale, therefore no statistical analyses were performed.
Clinical global impression (CGI) varied from 1 (very much improved) to 7 (very much worse). This information was collected only at week 12. Number and percentages of ''Global Improvement" categories are presented for each treatment group at week 12.
VNS satisfaction surveys consisted of 6 questions aimed at ascertaining the participant's acceptance and tolerability of the therapy and were collected in each treatment group at 12 weeks. These data are presented as a ratio of participants who responded positively to each question.

Results
Of the 15 sites recruiting for this study, 14 enrolled at least 1 patient. Following screening, 65 patients were enrolled, and 62 participants were both implanted and randomized into the three treatment groups (the ITT population) (Fig. 1). Of the three patients not included in the ITT population, two failed to receive implants due to insurance coverage, while one received implantation but was not set to the correct settings for their treatment group after randomization. This participant was included in the safety population. Five participants exited the study: One was lost to follow-up, three withdrew consent, and one was withdrawn by the site investigator after identifying the incorrect device was implanted (Model 102 implanted, study listed Models 103, 105, and 106 as permissi-ble). The average enrollment per site was 4.64 (median 3, range 1 to 12). Participant demographics are displayed in Table 2. There were no significant differences between the treatment groups at baseline with exception to the prevalence of anxiety disorders, which had a higher prevalence in Group C compared to Groups A and B (Group A/B/C: 13.6%/20%/50%). Generally, participants were white/Caucasian adult patients with DRE with an average age of 37.3 years (S.D. 12.0). Most participants had been living with epilepsy for over 10 years (15.5 years median time since diagnosis) and their seizures had failed more than 3 anti-seizure medications prior to enrollment (71% of participants). There were a greater number of female participants in this population (59.7% vs 40.3% male).

Clinical outcome -Cumulative achievement rate (CAR) and time until achievement (TUA)
Outcomes related to the effectiveness of each titration strategy in getting patients to dose can be found in Table 3. Overall, 39 participants (62.9%) were able to achieve the target settings by 12 weeks. Participants in Group A had the greatest likelihood to achieve the target settings by week 12, and 81.8% of participants in Group A achieved the target dose. Participants in Group B were least likely to achieve the target dose by week 12 (45%) and participants in Group C demonstrated a CAR between groups A and B (60%). While participants in Groups B and C were more likely to achieve the target dose later in the titration process, over 60% of participants in Group A achieved the target dose by week 8.
Time-to-achievement survival analysis was conducted and displayed in Fig. 2 and Table 4. Kaplan-Meier analysis revealed that the expected time-to-dose of Groups B and C were effectively identical, while participants in Group A were more likely to achieve the target dose by week 12. Participants in Group A were likely to achieve the target dose about 4 weeks earlier than participants using alternative titration protocols. The Kaplan-Meier median time to dose was estimated for Group A as 8.1 weeks, and estimated for Group C as 12.6 weeks. The model did not converge on an estimate for median time to dose for Group B as fewer than half of participants in this group achieved the target by 12 weeks.

Safety outcome -Device/Stimulation-Related adverse events (AE)
AE rates are presented in Table 5. Overall, 69 AEs were reported during the study, reported by 25 participants (40% of participants). AEs were reported at similar rates in each treatment group, with Groups A, B, and C reporting 25, 24, and 20 events, respectively. In each group, 33-45% of participants experienced an AE during the study. While Group C experienced a similar number of total AE as the other groups, it is worth noting that participants in Group C were the least likely to report AEs that were determined to be device/stimulation related. The most common AEs reported were consistent with real-world experience with VNS and included cough, voice alteration, general discomfort, and hoarseness.
Tolerability was assessed independently from AEs. AE reports were collected as the participant left the clinic following a titration event. Tolerability was assessed acutely as programming settings were adjusted in the clinic. Thus, AEs related to stimulation can be considered tolerability issues that did not fully resolve during the office visit. The fact that there were more tolerability issues than adverse events reported indicates that most tolerability concerns resolved during the office visit, and patients were able to proceed with titration on schedule. The presentation of tolerability concerns following VNS stimulation were similar to the most common VNS side effects and included general discomfort, cough, voice alteration, and hoarseness (most to least common). Participants in Groups A and C reported a similar number of tolerability concerns (36 and 34, respectively) while Group B reported a greater number of these tolerability issues (45 events).
One participant reported a Serious Adverse Event (SAE) during the study. The participant's reported event of non-healing wounds in the lower extremities was later assessed as related to severe obstructive peripheral vascular disease within the superficial femoral artery. As the event was not assessed to be related to VNS therapy, an unanticipated adverse event (UAE) was not recorded.

QOLIE-31
The QOLIE-31 was collected from all but 1 participant, who instead completed a QOLIE-48-AD due to their age. The QOLIE-31 outcomes are reported in Fig. 3. All treatment groups experienced a statistically and clinically significant improvement in the overall QOLIE-31 score versus baseline, although there were no significant differences between the groups.

CGI-I
CGI-I scores were collapsed from a 7-level scale into an 'improved' versus 'worsened' scale for ease of viewing and (Fig. 4). Data were missing in 3 participants coming from Groups A and B. Participants in Group C were least likely to demonstrate an improvement as perceived by the investigators and were more likely to worsen. Outcomes for Group A and B were generally similar, with 80% or more of participants improving compared to baseline.

Patient satisfaction
Patient satisfaction scores are described in Fig. 5. Participants in Groups B and C tended to report greater satisfaction with VNS therapy than participants in Group A. The most notable difference between groups was on the ''satisfaction with tolerability" query; all but 1 participant in Group C reported favorable responses, whereas 70% of participants in Groups A and B reported favorable responses. This outcome is further supported by the AE profile; participants in Group C participants reported the lowest number of device or implantation related adverse events (Table 5).

Discussion
The E-40 ''ASCEND" study is the first post-market double-blind randomized study of VNS therapy aimed to evaluate feasibility of achieving target device settings (OC = 1.5 mA, SF = 20 Hz, PW = 500 lsec and ON/OFF = 30 s/5min) within 12 weeks through three different titration schemes. An ultimate driving aspiration was to explore the practicality of reducing the time required to titrate VNS therapy to the target dose range given that evidence from clinical practice suggests that faster titration is associated with the highest likelihood of becoming a responder (>50% reduction in seizure frequency) quickly [6,7]. Yet, the same real-world data indicate that 50% of patients are not at the target dose within 12 months after implant [7]. The discrepancy may be indicative of considerable underutilization of clinical potential of VNS therapy, while about a third of patients with epilepsies are drug resistant and continue to suffer despite all pharmacological maneuvers [13]. Importantly, this suggests that even patients who are actively treated with VNS therapy may be underdosed and hence undertreated.
While a detailed dose-response relationship is established routinely for ASMs, it has only been strongly inferred for a VNS based on published trials [1,14,15] that were not designed to explore this challenge. Animal data suggest that only myelinated fibers A and B activation play a role in an antiseizure effect of VNS [16,17], and an astute modeling study [18] revealed ''sufficient vagal activation" with a combination of OC (0.75-1.5 mA) and PW (250 or 500 ls) settings in patients of age 12 or older. The combination of OC and PW defines the amount of charge delivered to the nerve fibers and indicates the likelihood of achieving an effective activation of the vagus nerve [18]. When thinking about attaining ''full", ''sufficient", ''effective" or ''optimal" activation of the vagus nerve, one must not forget that it requires selecting the combination of VNS settings for an individual patient with their unique tissue properties, electrode position(s), seizure type(s), comorbidities, ability, and readiness to tolerate stimulation. Thus, while insightful modeling informs us that we should use the indicated ranges of OC and PW to ensure the fiber activation, while maintaining patients' tolerability and preserving battery, clinicians are left with an individualized titration primarily guided with population outcomes and their modeled extrapolations until further translational research identifies a biomarker for sufficient vagal engagement that ensures an antiseizure effect without side effects. Like drug titration, the ultimate goal is the minimal amount of treatment, however defined, that produces a maximal lasting therapeutic   Table 5 Adverse events associated with titration according to each group titration schedule. Adverse events were collected as the patient left the clinic after a study-associated titration event, meaning that adverse events are limited to tolerability concerns that did not naturally resolve during the visit period.  effect with minimal side effects. Having multiple VNS parameters available makes it more challenging to define their optimal combination that would produce this kind of stimulation -a target dose. While the tolerability of a therapy at its target therapeutic dose is important to patient care, failing to achieve a therapeutic dose subjects a patient to potential risks without the opportunity for benefit. Specially for device-based therapies like VNS, the more threatening health risks like infection are derived from the surgical procedure, not the intensity of stimulation. Implanting VNS without following through to reach a therapeutic dose should be considered akin to subjecting a patient to a sham surgery, which is only considered ethical in the context of a clinical investigationnot clinical practice. We hope these findings will increase practitioners' zone of comfort and reduce the likelihood of encountering patients with uncontrolled epilepsy whose OC is below 1.0 mA even a few years after implantation (note: some patients may clinically respond below 1.0 mA, while others need currents well above 1.5 mA to respond [19]). However, it is worth mentioning that the parameter selections for titration of Groups B and C, namely lower frequencies and pulse widths, are widely believed to be more tolerable by patients. This anecdotal perception was supported by the study findings. The ASCEND study defines a therapeutic target for each VNS setting, but the therapeutic impact of specific signal frequency selection remains poorly explored in this population. Historically, VNS stimulation frequencies between 20 and 30 Hz have been used for VNS clinical trials in epilepsy and depression [15,[20][21][22][23][24]. The initial VNS clinical trials in epilepsy used 30 Hz stimulation frequency, while the clinical studies in depression have used 20 Hz. Over time, standard VNS frequencies for patients with epilepsy have also migrated to 20 Hz -so much so that it is now the default signal frequency for a new VNS device. This shift from 30 Hz to 20 Hz within the moderate frequency range has not been accompanied by any perceived impact on efficacy, at least at the population level, but has been accompanied by greater tolerability. In this study, patients in Groups B and C experienced lower stimulation frequencies, 5 Hz, at the beginning of their titration protocol and were later shifted to 20 Hz. While this decision was made to improve tolerability while the output current was increased, it is important to recognize that the decision to use signal frequencies outside of the 20-30 Hz range have not been thoroughly explored for the chronic treatment of epilepsy. In fact, a case report including human hippocampal EEG has shown decreases in interictal epileptiform discharges at 30 Hz stimulation, while VNS at the lower frequency of 5 Hz had the opposite effect [25]. Early studies of VNS in animal models of epilepsy indicated higher frequencies (>10, up to 20 Hz) to have a superior anticonvulsant effect when compared to lower frequencies (<10 Hz), but there is limited understanding of the impact of signal frequency selection (between 20 and 30 Hz) on epilepsy outcomes in humans [26]. More recent animal work indicates there may be a strong relationship between noradrenergic coeruleofugal signaling and VNS signal frequency [27].  While no UADEs were reported, stimulation AEs (28) outnumbered implantation AEs (17) as reported by only 40% of participants of ASCEND. Thus, most common adverse events encountered in our study (cough, voice alteration, hoarseness, general discomfort) parallel clinical practice experience where they are much more likely during the titration phase and often diminish with time [21,28]. Here, the stimulation associated AEs (patient's complaints reported on a departure from a clinic) were about 1/3 of the acute tolerability events (complaints reported during the visit which resolved by the time of departure), indicating that patients acclimate to the settings even during the office visit. Prior to the wider implementation of automated titration (''Scheduled Programming"), common clinical practice was to have patients sit in their waiting room for 30-60 min after a VNS adjustment to ensure their full comfort before departing from a clinic -especially if they live a long distance away. Assuming VNS is akin to drug titration, we would intuitively expect an inverse correlation at least between stimulation AEs (possibly all AEs) and achiever rate in this trial. Group B had the lowest achiever rate (45%), but the highest proportion of participants reporting stimulation AEs (33%). There was no such inverse association of dose achievement and stimulation AEs in groups A and C, and in fact a group A reported overall most AEs (25) and had the highest (81.8%) achievement rate. While we cannot explain these findings completely, it seems that a much finer capturing and analysis of AEs would be necessary to reveal if presence of any AE may have prevented further titration toward a target as opposed to only a moderate and/or severe AE doing so: This is where factors such as personal sensitivity and patient or caregiver motivation could play a role and affect the primary outcome. Clinical practice offers numerous experiences where psychological factors appear to be the key determinant of VNS settings and ultimate benefits it provides to the patients. We all have patients who accept a mild degree of hoarseness without any hesitation if they perceive any therapeutic benefits, but also others who may insist on an extreme step such as a device inactivation or exceptionally even its explanation without a clear externally verifiable reason. This is where practitioner's evidence-supported and experienceguided coaching of the patient is critical. However, anecdotal evidence suggests a considerably different degree of comfort with adjusting VNS between a novice and expert that could be improved with more evidence-based recommendations, including those ensuing from this study. Improvements in the various aspects of quality of life of patients with epilepsy (or depression) treated with VNS are solidly reported in the literature [29,30]. Anecdotally, ''more energy" or ''better mood" may be the first feedback to a clinician from a newly implanted patient. It may be worth noting that the authors would have a very hard time recalling a patient claiming the same after starting a new medication. Consequently, it is not surprising that all treatment groups experienced a statistically and clinically significant improvement in the overall QOLIE-31 score versus baseline.
CGI-I is a simple scale (0-7) on which patients rate subjectively a total clinical improvement from their baseline irrespective of their perception of its association with an intervention. In our study, about 80% of participants in Groups A and B reported an improvement compared to baseline, while participants in Group C were least likely to demonstrate an improvement and were more likely to worsen. Interestingly, the same number of participants in Group A and C reported ''no change". Considering that Group A includes the highest achievement rate (82%), and Group B the lowest (45%), it is safe to infer that CGI-I score is not driven by dosing alone. Curiously, Groups A and B had a similar number of total AEs, 25 and 24, respectively. The role and effect of subjectivity remains a complex concept that this study was not designed to address.
When it comes to patient satisfaction, participants in Groups B and C tended to report greater satisfaction with VNS therapy than those in Group A. The most notable difference between groups was on the ''satisfaction with tolerability" query, which had all but 1 participant in Group C reporting favorable responses, whereas 70% of participants in Groups A and B reported favorable responses. This outcome is further supported by the adverse events reported in the study, as Group C participants experienced the lowest number of device or implantation-related adverse events (Table 5). These findings suggest further complex associations between various subjective assessments where a convoluted interplay of personal, cultural, and societal elements may be in effect. Importantly, it is unknown what role a patient's disposition toward their disease and disease itself may be playing here.
The titration algorithm described in the manufacturer's labeling, which is also pre-programmed into the guided programming and scheduled programming features on the most recent VNS generator SenTiva, is the Group A protocol. This study suggests that >80% of patients could achieve the target settings in 3 months if this protocol is followed. Yet, product surveillance indicates that 50% of patients in the US are not at the target dose within 12 months after implant [7]. This is a solvable problem, and scheduled programming can even offload the burden of the biweekly inoffice visit cadence required to follow this protocol precisely. Importantly, participants in Group B and C were titrated at a lower PW and a very low SF that is practically not used in clinic and were forced to titrate initially all the way to 2.5 mA before lowering the OC to 1.5 mA and changing other parameters. It is possible that a different titration scheme that allowed for more settings to be changed at once -rather than OC first, then PW, then other settings -would have allowed these participants to be titrated faster. In short, it is critical that practitioners realize that a VNS titration timeline can be shortened drastically and safely to decrease a seizure burden and preserve a brain reserve, while improving various aspects of patients' lives.
Although prospective, blinded, and randomized, our study has a few limitations: 1. Group size was not a product of power calculation but rather an empiric heuristic; 2. except for a Group A protocol for which preexisting evidence from epilepsy field was available, other groups' titration schemes were more convenience choices based on different contexts not reflective of clinical epilepsy practice; 3. subjective scales were used, and their results may be affected by many factors that cannot be accounted for; and 4. we do not have a befitting account for a reason(s) that prevented reaching target parameters among some of our participants.

Conclusion
In short, this study proves that achievement of an adequate dose of VNS therapy, a dose that is associated with ideal therapeutic outcomes in most patients [6], is possible in 3 months or less for most patients. Patients who receive regular increases to VNS intensity (output current increasing at a stable pulse width) while other stimulation parameters are held constant at the default settings are most likely to achieve this target dose of 1.5 mA, 500 ls, 20 Hz, and 10% duty (30 seconds ON, 5 minutes OFF). This titration protocol, known in VNS programming parlance as ''the standard protocol", can be automatically implemented on more recent VNS models through scheduled programming.

Author contributions
AB and RV were responsible for the initial draft of the manuscript, with PA and SB offering the first critical review and addi-tional content. The ASCEND Study Group offered critical input during a second round of review, with SB offering commentary that went beyond what was offered by others in the group. All authors were responsible for recruitment and data collection for the ASCEND study, with SB's center making the greatest contribution to the study population. Authors associated with the study sponsor were responsible for study design and execution, databasing, data monitoring, and other administrative activities.

Declaration of Competing Interest
The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: AB currently receives clinical research funding from Liva-Nova USA Inc. for participation in the CORE-VNS Registry as a principal investigator. RV is an employee of LivaNova USA Inc. and holds stock or stock options. SB is a scientific consultant and serves on the speakers' bureau for LivaNova USA, Inc. PA has no conflicts of interest to report. LivaNova USA Inc. is the manufacturer of the VNS Therapy System and was the sponsor of the ASCEND study.
We confirm that we have read the Journal's position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.